Prof. Conte wrote the following hallowed words in a Teams chat window on September 22nd, 2022:
I'm going to assume you guys want to publish. I've been distracted by my day job and not explained my strategy (advice) that I have given past students. So tossing the ADR hat aside...
Aim for top conferences only
Even if you do not get accepted, the reviews will help you get in the next time. What our our top conferences? ISCA, HPCA, MICRO are the core conferences. You should know the deadlines:
- http://www.wikicfp.com/cfp/program?id=1683&s=ISCA&f=International%20Symposium%20on%20Computer%20Architecture
- http://www.wikicfp.com/cfp/program?id=2052&s=MICRO&f=International%20Symposium%20on%20Microarchitecture
- http://www.wikicfp.com/cfp/program?id=1220&s=HPCA&f=High-Performance%20Computer%20Architecture
Write the narrative part of the paper first; Write it well
A paper shouldn't just have the technical information right, it shouldn't just have all of the details in there, it doesn't just need to have all of the experiments done and the results explained (note, having all that is not optional), but also a paper needs to be interesting, easy to read, and informative.
Plan ahead.
A good paper is a huge effort that takes time. There is no substitute for putting in the time. It helps to team up on papers. Austin, you can help Pulkit on his, and vice versa, and also the same with other members of the group.
Keep me in the loop. Remember I know how to do this. 100+ published papers.
You will get discouraged from time to time. Psych yourself up for the long haul. This is not going to happen overnight. You might have to do "the loop" (ISCA->HPCA->MICRO->...) several times with a submission to get it published.
So, commit yourselves to this. It is not easy, it's hard, it will take a lot of work. (But if Ph.D.'s were easy, everyone would have one and they'd be worth shit.)
Print that out. Hang it on your fridge. Put it across from your toilet so you can read it when you take a crap.
Pulkit replies,
For #2, do you mean that we should write the motivation and "value" of the work first? And then fill in the gaps with technical shit and numbers to back up the "value"?
Tom responds,
Not first— in-addition-to. It needs to be a fun read — easy to read, interesting, informative. It must answer the first four questions of the Heilmeier Catechism:
- What are you trying to do? Articulate your objectives using absolutely no jargon.
- How is it done today, and what are the limits of current practice?
- What is new in your approach and why do you think it will be successful?
- Who cares? If you are successful, what difference will it make?
In response to murmurs in the crowd about compilers, Tom clarifies,
For languages and compilers, the chief conferences are a little different. They are PLDI, LCxx (there are several) that are run by SIGPLAN (the same folks who run PLDI), and then CGO for compiler optimization.